Resource track
Scope
This MELBA Resource track is dedicated to resources that can facilitate i) reproducible research within the field of biomedical imaging and ii) fair benchmarking of machine learning algorithms.
MELBA Resource publications are articles systematically describing publicly available resources, such as i) research datasets and ii) open-source software tools, that reside at the intersection of biomedical imaging and machine learning.
Our aim is to enhance the sharing and reuse of resources, encourage their broader dissemination and repurposing, as well as acknowledge those who contribute by sharing.
Format of Resources
The sections of a MELBA Resource manuscript that potential submissions should include are:
- Title
- Abstract
- Background
- This section should provide the background and motivation for the specific “resource”.
- The gap/need in the literature that makes the presented “resource” valuable for ML/AI projects should be indicated.
- Reference to similar “resources”, further highlighting the unaddressed gap or complementing the current “resource”, should also be reported here.
- Summary
- This section should summarise the vision, mission, scope, and objectives of the “resource”.
- Who is the target audience? Who does it serve?
- The readiness of the “resource” for use in a medical imaging ML/AI project is expected to be highlighted here. For example, is the presented dataset AI-ready? Does use of the presented tool require programming knowledge? Is the presented tool executable in a low-/zero-code principle?
- Discussion
- Resource Availability
- Summary statement
- This section should summarise the importance of the resource (e.g., new datasets for underserved populations), in not more than 4 sentences.
- Data/Code Location
- Specific link should be added here.
- Examples of repositories are given at the Public Repositories section below.
- Potential Use Cases
- This section should indicate the potential ML/imaging use cases that the authors envision their resource facilitating and the specific potential clinical domain(s).
- Licensing
- Clearly describe the licensing terms.
- If there is an existing licence that is followed this could just be a single sentence.
- Ethical Considerations
- Language referring to associated documentation, e.g., IRB, Informed Consent.
- For human data collection this section must describe the consent or opt-in/out process.
- Methods
- This section should highlight either (for ‘Data Resource’ papers) all methods/techniques/pipelines that the authors used to prepare the presented data, or (for ‘Tools Resource’ papers) that were included in the presented tool.
- Data Details
- This section should indicate inclusion criteria, as well as describe any associated demographic, clinical, molecular, and other related details.
- Methods Used for the Data Creation
- This section is only relevant for ‘Data Resource’ papers.
- All methods and equipment that were used from data acquisition to final processing should be described in detail here.
- Describe the AI-ready state of the dataset. Is it ready to be ingested by a DL algorithm? If not, what are the potential steps that one needs to take for this?
- Software stack
- This section is only relevant for ‘Tools Resource’ papers.
- A textual description and a graphical illustration must be included here.
- Flowchart(s) illustrating the tool’s process flow
- This section is only relevant for ‘Tools Resource’ papers.
- For example, this flowchart could be indicative of the ML training procedure in this tool.
- This flowchart does not necessarily need to be a formal Data Flow Diagram
- Functionalities (For Tools Resource papers only)
- This section is only relevant for ‘Tools Resource’ papers.
- Validation
- Description of approaches ensuring quality of data
- This section is only relevant for ‘Data Resource’ papers.
- Subsections included here should be corresponding to the “Methods Used for the Data Creation” subsection above.
- Experimental Design & Results
- This section is only relevant for ‘Tools Resource’ papers.
- Indications of the reproducibility of the code.
- CI/CD - Unit Test code coverage
- Example data from published or unpublished studies
- Acknowledgements
- Funding and conflicts of interest.
- References
Submission Considerations
MELBA Resource publications are intended to supplement conventional research publications. The release of MELBA Resource articles will not be regarded as diminishing the originality of subsequent MELBA manuscript submissions. Authors are encouraged to consult with editors from other journals to ensure that the publication of a MELBA Resource does not compromise the submission of related research articles to those journals.
Authors may also submit a MELBA Resource manuscript detailing data/tools previously made publicly available or analysed in prior research articles, provided there is no existing publication specifically describing the resource. In case of an existing publication, a new publication might be submitted only if it significantly enhances the resource's reuse potential or presents a substantial extension of the previously published resource.
Any pre-existing publication, or manuscript under consideration or in press at any journal, relevant to the MELBA Resource article must be mentioned and explicitly cited in the submitted manuscript, as well as discussed in the accompanying cover letter.
Authors submitting a MELBA Resource manuscript are required to deposit their data/tools in a suitable external public repository to ensure the discoverability, utility, reproducibility, and reusability of the resource. Sole use of a personal or institutional website is not considered adequate. Authors should furnish their materials at a level that facilitates broad utility, encompassing both computed or curated data and observed (source) data.
Should the resource not be publicly accessible at the time of submitting a MELBA Resource manuscript, authors are required to furnish secure links and/or passcodes. This allows the editorial team and reviewers to confidentially access and assess the data. Authors are advised against providing personal login details and are not allowed to track individual logins that would compromise the anonymous reviewer evaluation. The editorial team and reviewers commit to treating the manuscript and its associated data/tools with strict confidentiality, utilizing them solely for the purpose of manuscript evaluation prior to publication.
In the event that the data/tools are not accessible at the original repository, authors are obligated to upload the data to another repository and issue an update to the original MELBA Resource article. If the authors fail to respond accordingly, the editorial team reserves the right to retract the publication.
MELBA maintains the right to decline a Resource article post-acceptance, if significant issues with the scientific content or violations of our publishing policies emerge.
Public Repositories
The MELBA Resource expectations for the repository of the data/tools should enable the long-term preservation and public access to the uploaded data/tools, without paywalls, and support open licenses (such as CC-BY). Example repositories are:
- Synapse
- The Cancer Imaging Archive (TCIA)
- Neuroimaging Informatics Tools and Resources Collaboratory (NITRC)
- Zenodo
- EBRAINS
- National Addiction & HIV Data Archive Program (NAHDAP)
- Dryad Digital Repository
- Science Data Bank
- NeuroMorpho.org
- G-Node
- Harvard Dataverse
- UK Data Service
- Image Data Resource
- ClinicalTrials.gov
- OpenNeuro (formerly OpenfMRI)
- SICAS Medical Image Repository (formerly Virtual Skeleton Database)
- National Database for Autism Research (NDAR)
- Open Science Framework
- National Database for Clinical Trials related to Mental Illness (NDCT)
- PhysioNet
- figshare
- Research Domain Criteria Database (RDoCdb)